Sharpness of the phase transition in percolation models

Michael Aizenman, David J. Barsky

Research output: Contribution to journalArticlepeer-review

231 Scopus citations

Abstract

The equality of two critical points - the percolation threshold pH and the point pT where the cluster size distribution ceases to decay exponentially - is proven for all translation invariant independent percolation models on homogeneous d-dimensional lattices (d≧1). The analysis is based on a pair of new nonlinear partial differential inequalities for an order parameter M(β, h), which for h=0 reduces to the percolation density P - at the bond density p=1-e in the single parameter case. These are: (1)M≦h∂M/∂h+M2+βM∂M/∂β, and (2) ∂M/∂β≦|J|M∂M/∂h. Inequality (1) is intriguing in that its derivation provides yet another hint of a "φ{symbol}3 structure" in percolation models. Moreover, through the elimination of one of its derivatives, (1) yields a pair of ordinary differential inequalities which provide information on the critical exponents {Mathematical expression} and δ. One of these resembles an Ising model inequality of Fröhlich and Sokal and yields the mean field bound δ≧2, and the other implies the result of Chayes and Chayes that {Mathematical expression}. An inequality identical to (2) is known for Ising models, where it provides the basis for Newman's universal relation {Mathematical expression} and for certain extrapolation principles, which are now made applicable also to independent percolation. These results apply to both finite and long range models, with or without orientation, and extend to periodic and weakly inhomogeneous systems.

Original languageEnglish (US)
Pages (from-to)489-526
Number of pages38
JournalCommunications In Mathematical Physics
Volume108
Issue number3
DOIs
StatePublished - Sep 1987
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • Statistical and Nonlinear Physics
  • Mathematical Physics

Fingerprint

Dive into the research topics of 'Sharpness of the phase transition in percolation models'. Together they form a unique fingerprint.

Cite this